摘要 |
Methods for dimensionality reduction of large data volumes, in particular hyper-spectral data cubes, include providing a dataset &Ggr; of data points given as vectors, building a weighted graph G on &Ggr; with a weight function w&egr;, wherein w&egr; corresponds to a local coordinate-wise similarity between the coordinates in &Ggr;; obtaining eigenvectors of a matrix derived from graph G and weight function w&egr;, and projecting the data points in &Ggr; onto the eigenvectors to obtain a set of projection values &Ggr;B for each data point, whereby &Ggr;B represents coordinates in a reduced space. In one embodiment, the matrix is constructed through the dividing each element of w&egr; by a square sum of its row multiplied by a square sum of its column. In another embodiment the matrix is constructed through a random walk on graph G via a Markov transition matrix P, which is derived from w&egr;. The reduced space coordinates are advantageously used to rapidly and efficiently perform segmentation and clustering.
|